/A practical guide for PT practices
AI in Physical Therapy: Clinical and Administrative Use Cases
Where AI fits in a physical therapy practice, what needs human review, and how to tell whether it is helping your clinic.
What is physical therapy AI?
Physical therapy AI is software that uses artificial intelligence to support work in a PT practice. That can mean answering a patient's call, drafting a treatment note, or helping a clinician interpret movement data. The useful question is which task the tool can complete reliably, in your setting, with appropriate oversight.
Separate the evaluation into two areas: administrative tools that help patients access care, and clinical tools that support the care team. Success for a scheduling agent means an accurate appointment and a clear patient record. Success for a clinical tool needs evidence relevant to its intended use and the patients it serves.
APTA recognizes potential for AI to expand access, support care delivery, and reduce administrative burden in its policy on digital health and AI in physical therapist practice. That potential is a starting point for evaluation, not proof of a particular product's results.
Physical therapy AI use cases at a glance
| Workflow | What AI can help with | What your team owns |
|---|---|---|
| Patient calls and texts | Answer routine questions, follow up with leads, and route requests. | Define which conversations go to staff. |
| Scheduling | Book evaluations and follow-ups; handle appointment changes. | Set provider, location, visit type, and capacity rules. |
| Insurance verification | Gather coverage, cost share, visit limits, and benefit details. | Review unclear results and keep the source evidence. |
| Prior authorization | Organize documents, submit requests, and track status. | Review clinical documentation and payer decisions. |
| Documentation | Draft a note from an encounter or clinician input. | Check accuracy and approve the final clinical record. |
| Clinical support | Assist with movement measurement or review patient progress. | Evaluate evidence and make diagnosis and treatment decisions. |
These are product categories to evaluate. A single tool may cover only part of a workflow.
AI calls and scheduling for physical therapy
Patient access is a concrete place to evaluate AI. A caller may need a new evaluation, a different appointment time, or help finding the right location. Physical therapy voice AI can gather that request, check the clinic's rules, and complete supported scheduling work while keeping staff available for exceptions.
For example, an after-hours caller requests an evaluation. The workflow should identify the right location and provider, check the appointment type and duration, and save a booking in the EMR. If the request needs a person, the handoff should explain what the patient asked and what remains unresolved.
In a demo, test the full task: an unavailable provider, a caller who changes their mind, a reschedule, and a clinical question. Check the resulting appointment and patient record. A pleasant conversation is only one part of a completed workflow.
Shasta's AI phone calls for physical therapy clinics support inbound calls, texts, new-lead outreach, and scheduling in the connected EMR. You can begin with after-hours or overflow coverage and define when staff should take over.
Clinical support and AI documentation
Drafting and reviewing notes
AI documentation tools can help turn encounter information into a draft note. APTA's introduction to its ambient scribe practice advisory describes tools that capture, transcribe, and summarize patient-provider interactions. Evaluate how a tool distinguishes speakers, handles corrections, and fits your documentation process.
Plan for clinician review before a draft becomes the final record. Test whether the note reflects what actually happened, including measurements, patient statements, and the plan of care. Measure editing time alongside drafting time so the evaluation captures the full workload.
Movement analysis and decision support
Clinical AI categories also include tools designed to assist with movement measurement or summarize patient progress. Ask for validation that matches the proposed task, patient population, and setting. Review how the tool reports uncertainty and how a therapist can inspect or correct its output before using it to inform care.
Shasta's agents described here focus on patient access and administrative work. AI scribes and clinical analysis tools are separate categories; this guide does not present them as Shasta capabilities.
Safety, privacy, and human oversight
Set review rules before connecting a tool to patient information. HHS explains that a cloud provider handling electronic protected health information on behalf of a covered entity generally acts as a business associate, with a business associate agreement and other HIPAA obligations. See HHS guidance on HIPAA and cloud computing when reviewing the proposed arrangement.
- Data access: Which records can the tool read or change, who can review them, and how are permissions removed?
- Data handling: Where do recordings, transcripts, and notes go? What are the retention, deletion, and model-training policies?
- Patient communication: How are patients informed about AI, recording, and outreach? Review applicable requirements with your privacy or compliance lead.
- Escalation: What happens with a clinical question, an urgent concern, an identity mismatch, or a failed integration?
- Accountability: Can staff see the source information, the action taken, and the person responsible for the next step?
For Shasta's security information and review materials, visit the Trust Center.
How to measure physical therapy AI ROI
Record a baseline before launch. For calls, track answer rate, completed bookings, staff handoffs, and errors that require correction. For insurance work, track completion before the visit and staff review time. Compare similar periods and account for changes in call volume, staffing, and appointment availability.
An illustrative monthly calculation
If a clinic saves 40 staff hours at a loaded cost of $30 per hour, the capacity value is $1,200. If 10 additional completed visits each contribute $60 after variable costs, they add $600.
$1,200 + $600 − $1,000 in total monthly costs = $800 in net monthly value.
That is an illustrative 80% return on the $1,000 cost. These are hypothetical inputs, not Shasta pricing or promised results. Include software, usage, implementation allocation, training, and ongoing review in your costs.
Saved time is capacity unless it reduces spending or produces measurable additional work. Count completed visits rather than bookings, subtract staff correction time, and avoid counting the same benefit twice. Track patient experience alongside the financial estimate.
Start with a workflow you can verify
- Pick a specific bottleneck. Define the task, the systems involved, and what counts as done.
- Test realistic cases. Include ordinary requests, missing information, changes mid-conversation, and cases that need a person.
- Agree on the handoff. Assign ownership for exceptions and confirm staff can see what happened.
- Review the pilot. Compare completion, accuracy, review time, and patient feedback against your baseline before expanding.
For an example of keeping administrative workflows connected to clinic systems, read how Enduring Performance kept calling and insurance verification running through an EMR transition.
Common questions about AI in physical therapy
What is physical therapy AI used for?
Physical therapy AI can support administrative work such as calls, scheduling, insurance verification, and authorization tracking, as well as clinical work such as documentation and movement analysis. Capabilities differ by product. Each workflow needs clear limits and an accountable staff member or clinician.
Can AI replace a physical therapist?
This guide treats AI as a support tool. Patient evaluation, diagnosis, treatment planning, and clinical judgment remain with the qualified care team. A tool that handles calls or drafts notes should not be assumed to provide clinical care.
How is physical therapy voice AI different from an AI scribe?
Voice AI for patient access handles phone conversations and administrative actions such as scheduling. An AI scribe helps draft clinical documentation from an encounter. They address different workflows and need different review and integration requirements.
Where should a PT clinic start with AI?
Choose a specific workflow with a measurable baseline and a clear handoff to staff. After-hours calls, appointment changes, or benefit checks can be practical starting points. Test routine cases and exceptions before expanding coverage.